AI News Roundup: Pentagon Freezes Out Anthropic, KKR Bets $10B on AI Infrastructure, EU AI Act Talks Stall

Pentagon Signs Classified AI Deals With Eight Companies — Anthropic Left Out

The Department of Defense announced agreements with eight major technology companies to deploy AI tools on its highest-classification networks. SpaceX, OpenAI, Google, NVIDIA, Microsoft, AWS, Reflection, and Oracle will gain access to the Pentagon’s Impact Level 6 and Impact Level 7 environments — the two most sensitive tiers of classified infrastructure.

Notably absent: Anthropic, which the Pentagon designated a “supply-chain risk” in March after the company refused to grant unrestricted access to its Claude models for use in autonomous weapons and mass domestic surveillance. Anthropic is currently suing the DOD over the label, calling it retaliatory. Pentagon CTO Emil Michael framed the exclusion in pragmatic terms, saying the department has come to recognize that being reliant on any one partner is irresponsible.

A path forward may still exist. Trump’s chief of staff Susie Wiles and other senior officials met with Anthropic CEO Dario Amodei at the White House on April 17, and Trump told CNBC a deal was “possible.” But for now, the world’s most safety-focused AI lab remains locked out of the world’s largest defense budget.

KKR Launches $10 Billion Helix Digital Infrastructure

Private equity giant KKR secured more than $10 billion to launch Helix Digital Infrastructure, a new company that will design, build, own, and operate the physical layer of AI — data centers, power generation and transmission, and fiber connectivity. Former AWS CEO Adam Selipsky will serve as CEO and chairman.

The thesis: the real bottleneck in AI isn’t models or chips but the infrastructure around them. Helix will partner directly with hyperscalers so they can keep training and deploying models at scale without getting stuck in construction delays, permitting backlogs, or power shortages. The initiative has already attracted commitments from sovereign wealth funds and strategic partners, with additional fundraising anticipated.

Anthropic Surpasses OpenAI in Revenue at $30B Run Rate

Anthropic’s annualized revenue run rate has crossed $30 billion, up from roughly $9 billion at the end of 2025 — more than tripling in about four months. The company now counts over 1,000 enterprise customers spending more than $1 million annually, a figure that has doubled since February.

The milestone puts Anthropic ahead of OpenAI’s most recently reported $25 billion ARR, though OpenAI’s chief revenue officer disputed the comparison, arguing Anthropic’s number is inflated by roughly $8 billion because it reports gross revenue from AWS and Google Cloud channel sales rather than netting out partner cuts. Regardless of accounting methodology, the growth trajectory is staggering and underscores how deeply enterprise demand for Claude has accelerated in 2026.

OpenAI Ships GPT-5.5 With Strongest Agentic Capabilities Yet

OpenAI released GPT-5.5 and GPT-5.5 Pro to API users and ChatGPT subscribers, calling it the company’s most capable model to date. The gains are concentrated in agentic coding, computer use, knowledge work, and early scientific research — areas where progress depends on reasoning across long context and taking autonomous action over time.

GPT-5.5 matches its predecessor’s per-token latency while delivering significantly higher intelligence, and it uses fewer tokens to complete the same Codex tasks, making it both more capable and more efficient. OpenAI says you can now hand the model a messy, multi-part task and trust it to plan, use tools, check its work, and keep going without step-by-step supervision.

EU AI Act Reform Talks Collapse After 12-Hour Marathon

The European Parliament and Council failed to agree on amendments to the EU AI Act after a grueling 12-hour negotiation session that began April 28 and ended in the early hours of April 29. The sticking point: whether AI embedded in products already covered by EU safety rules — machinery, toys, medical devices — should get an exemption from the Act’s high-risk requirements.

The stakes are significant. The AI Act’s core obligations for high-risk systems take effect on August 2, 2026 — just three months away. The entire purpose of the Digital Omnibus for AI is to push that deadline to December 2027 for stand-alone systems and August 2028 for regulated products. A follow-up trilogue is scheduled for approximately May 13. If talks stall again, the original August deadline stands and thousands of companies deploying AI in Europe will face compliance requirements many say they aren’t ready for.

Standard Intelligence Raises $75M for “Computer Use” AI

Six-person startup Standard Intelligence raised $75 million from Sequoia and Spark Capital, with backing from angel investors including Andrej Karpathy. The company has built FDM-1, a foundation model optimized for interacting with applications through their graphical interfaces, trained on a dataset of 11 million hours of screen-recording footage.

FDM-1 can perform tasks ranging from scanning software for security vulnerabilities to operating computer-aided design tools. In one demonstration, the model learned to drive an autonomous vehicle simulation after just one hour of fine-tuning — a striking example of how quickly these models can adapt to new visual interfaces.